An L2 Pronunciation Judgment Task
Adolfo Ausín and Megan Sutton
Michigan State University
1. Introduction
Elliott (2003) writes:
“[w]ith the advent of more communicative approaches to language learning and instruction in the
1980s, explicit instruction in target language pronunciation became a thing of the past and was
generally limited to the preliminary chapters of language textbooks or found in the back pages of
the textbook appendixes.” 1
As a remedy to what we might call the pronunciation deficit of the communicative approach, many
undergraduate Spanish programs across the US offer a course in phonetics and pronunciation in their
third or fourth year. An indication of the interest in these types of courses is the number of phonetics /
pronunciation textbooks that are available on the market, such as Guitart (2004), Hammond (2001),
Quilis and Fernández (2003), Schwegler, Kempff and Barrutia (2007), Spicher, Sweeney and Pelayo
Coutiño (2007), Stokes (2005), and Teschner (2000). Of course, not all these textbooks have the same
emphasis, and there is great variation with respect to the degree of sophistication of the phonetic and
phonological theory that one can find in these textbooks. Nevertheless, all of them have the practical
purpose of improving the students’ pronunciation.
At the same time that formal training in phonetics and pronunciation has found a way into the
Spanish curriculum, there has been an increasing interest in finding out whether such an addition is
adequate. One thing is to realize that the communicative approach to language teaching is not very
effective at getting the students to pronounce correctly (after all, precise pronunciation is seldom an
impediment for successful communication). It is a very different thing is to claim that formal teaching
in phonetics can have an impact on improving the students’ pronunciation. For this reason, there have
been studies devoted to test whether formal instruction is an effective way of improving the students’
pronunciation.
Although the students in a phonetics / pronunciation course might learn about the different modes
of articulation or how the concepts of phonemes and allophones help us understand the differences
between [s] and [z] in Spanish and English, the question is whether students actually improve in their
practical knowledge of the language; that is, whether they put the phonetic theory to use. There are
impressionistic data that seem to indicate that the students’ pronunciation of Spanish does not improve
after a phonetics class. If there is no improvement at all, it would be difficult to justify the need for this
type of course, although not impossible, since one could argue that it is important that students are
aware of the phonological properties of the language even if they cannot put them to practice.
Many studies that test the effectiveness of formal phonetics seem to indicate that formal training
in phonetics has significant positive impact on the students’ pronunciation. (See, for example,
González-Bueno (1997), Lord (2005), Gaff and Loewen (2007), and Elliott (2003), which provides an
overview of L2 studies in the acquisition of phonology.) A notable exception to this is Sutter (1976),
who found that having formal training in pronunciation did not correlate with better pronunciation.
1
Similar remarks can be found in Pennington and Richards (1986), Terrell (1989), Elliott (1995), Cenoz and
Lecumberri (1999), and Lord (2008) among others.
© 2010 Adolfo Ausín and Megan Sutton. Selected Proceedings of the 12th Hispanic Linguistics Symposium,
ed. Claudia Borgonovo et al., 234-245. Somerville, MA: Cascadilla Proceedings Project.
235
2. Testing linguistic competence
In this paper we present an experiment designed to test whether a phonetics class actually makes
an impact on the linguistic competence of the enrolled students. The novelty of our experiment resides
in the way we test our subjects’ knowledge of Spanish pronunciation. In most experiments that we are
aware of, the subjects produce some speech samples which are then rated by native speakers of the
target language, or are measured and the results compared with those of native speakers. For instance,
Colantoni and Steele (2006) and Face (2006) had their subjects read a passage, recorded their passages,
analyzed some relevant tokens, generated relevant spectrograms and then measured features such as
stop length and voicing or the presence of a brief closure. On the other hand, Gaff and Loewen (2007)
recorded their subjects performing a series of tasks (free response picture description, paragraph
reading, sentences reading, word list reading) and then raters judged the subjects’ utterances.
Although this way of testing pronunciation seems ideal, it frequently proves to be impractical.
Since it requires a considerable amount of time to collect and analyze the data, frequently the number
of subjects under investigation has to be kept small, which might end up making the (statistical)
analysis of the data more problematic. Also, the reliance on subjective raters adds another level of
complexity which again might make the data analysis more problematic and cumbersome. For these
reasons we decided to look for an alternative way of determining the linguistic competence of the
subjects, and we decided to assess perception, not production.
There have been many studies that have focused on the perception abilities of second language
learners. For example, Zampini (1998) measured the VOT perceptual boundaries between /p/ and /b/.
She modified the VOT of /p/ and /b/ in increments of 5 milliseconds and used this VOT continuum to
determine at what point the subjects placed the difference between /p/ and /b/. (She found no
correlation between perception and production of /p/ and /b/.) Flege and MacKay (2004) tested the
ability of native Italian speakers to perceive English vowels correctly. They used two different testing
strategies but in both the subject had to listen to a series of English minimal pairs (/bit/ vs. /bɪt/).
Nibert (2006) presents what could be considered a Truth Value Judgment Task (TVJT). She is trying
to see whether non-native speakers of Spanish can use the intonational clues that distinguish the two
interpretations of sequences like “lilas y lirios amarillos” in the same way native speakers do.
(1) Example of a minimal group, using the text “lilas y lirios amarillos”
Tonal structure of the intonation contours
Possible meanings:
a. [[lilas] [y lirios amarillos]]
→ A. ‘lilacs and yellow irises’
HL- L%
b. [[lilas y lirios] [amarillos]]
→ B. ‘yellow lilacs and yellow irises’
HL- L%
c.[[lilas y lirios amarillos]]
→ A or B. No disambiguating medial H- phrase
L- L%
accent is present.
Nibert presented the above utterances and asked her subjects whether each “utterance refers to two
types of flowers: lilacs (=lilas) of some unknown color, and yellow irises (=lirios).” Finally, Grenon
(2006) used a perceptual discrimination test to test whether Japanese speakers could discriminate
certain contrasts present in English but not in Japanese. Some of the nonsense words that she used
were /sɪd-θɪd/ and /zug-ðug/.2
In all of these perception studies the subjects “perceive” correct grammatical sequences, that is,
fragments that were pronounced correctly.3 Now we would like to turn our attention to two studies that
followed a different approach: Sakow and McNutt (1993) and Altenberg (2005).
2
See Flege (2003) for more studies on L2 perception (and production) as well as for theoretical considerations
that fall beyond the scope of this article.
3
In the Zampini study the subjects were not listening to normal fragments in the sense that they had been
artificially modified by the researcher, but the subjects were under the assumption that they were “normal”
fragments. In the Grenon study, even though the words were made up, they were still pronounced correctly
because they conformed to the phonological requirements of English phonology, and therefore, were grammatical.
236
Sakow and McNutt (1993) tested how well native speakers of Japanese and Korean could
distinguish between correct and incorrect pronunciations of /r/ in English. They tested both external
and internal perception. In external perception the utterances being judged were recorded by native
speakers of English. In internal perception, subjects judged their own utterances that they had
previously recorded. They found that subjects were better at external perception.
Altemberg’s (2005) goal was to determine the knowledge that native speakers of Spanish learning
English had regarding what constitutes a possible consonant cluster in English. One of the tasks that
she used, a Metalinguistic Judgment Task, asked subjects to judge the acceptability of written nonwords, such as “slad” and “fnod” as possible words in English. That is, she used an explicit
grammaticality judgment task to determine the subjects’ phonological knowledge.
The testing techniques that were used in Sakow and McNutt (1993) and Altenberg (2005) could be
described as a grammaticality judgment task (GJT), since the subjects were asked to determine
whether certain stimuli were acceptable or not in the target language. The use of grammatical
judgments in SLA has been debated (see Sorace (1996)), however Chaudron (2003) notes that their use
is standard in L2 syntactic and semantic research. As an example of the syntactic application of the
GJT, consider Zyzik (2006). In her study, Zyzik focuses on the status of null objects in learners of
Spanish. One of the tasks that her subjects performed was to determine whether the sentences that
appear in (2) are acceptable or not.
(2) a.
b.
¿Qué hiciste con el pastel? Lo puse en el refrigerador.
‘What did you do with the cake? I put it in the fridge.’
*Ellos compraron la cerveza y pusieron en el refrigerador.
‘They bought the beer and put in the fridge.’
Zyzik used her subject responses to these sentences to determine the (syntactic) knowledge that
her subjects had about this type of constructions. Similarly, Pérez-Leroux, Majzlanova, and SánchezNaranjo (2003) use a grammaticality judgment task to determine their Spanish L2 learners’ knowledge
of the semantic factors involved in the choice between preterite and imperfect in Spanish. They asked
their subjects to rate semantically well- and ill-formed sentences like the following:
(3) a.
b.
Hice la tarea pero no me dejaron terminarla. (semantically ill-formed)
Hacía la tarea pero no me dejaron terminarla. (semantically well-formed)
‘I did/was doing the homework but they didn’t let me complete it.’
Pérez-Leroux et al. use their subjects’ ratings for sentences like the previous ones to determine
their subjects’ knowledge of the semantic features involved in the correct choice of preterite vs.
imperfect.
As can be seen, the GJT has been used to assess the linguistic knowledge in the realms of syntax,
semantics, and to a lesser degree phonetics / phonology. For our experiment, we decided to follow
Altermberg’s and Sakow and NcNutt’s strategy and tested our subjects’ knowledge of Spanish
phonetics by using a pronunciation GJT. The general idea was to offer subjects good and bad
recordings of Spanish words, and ask the students to rate them, to see whether they were able to tell
the good recordings from the bad ones.
One immediate practical advantage of our approach was the simplification of the analysis of the
subjects’ responses. As mentioned earlier, in a pronunciation production task, after the subject
produces the relevant samples, the researcher has to spend time and energy analyzing the data, in a
fairly complicated process of analyzing what the subjects have produced. Furthermore, the complexity
of the process increases the chances of error and decreases the possibility of studying a large subject
population. In a pronunciation grammaticality judgment task, the analysis of the subject’s response is
much simpler, the possibilities of error are virtually zero, and the analysis of large subject population is
completely plausible.
237
3. The experiment
3.1. Subjects
The test (a pre- and a post-test) was initially administered to all the students enrolled in two
sections of a Spanish pronunciation and phonetics class at an American university. The class is
described in the catalog as a “[p]honetic description of the sound system of Spanish, developed
linguistically and applied to the improvement of pronunciation and spoken Spanish.” Students take this
class approximately during their third year of college-level instruction.
The class follows somewhat closely the structure of the textbook, Stokes (2005). The students are
introduced to the spelling, phonetic and phonological systems of Spanish. In general a typical class
period includes a review of daily homework (which might involve some phonetic transcription),
lecture by the instructor4 about some phonetic topic, and some oral practice exercises that cover the
day’s material. Even though the textbook comes with an audio program CD, it was rarely used. At the
end of the semester there was an oral project where students had to make a recording of the poems in
the book / CD.
A total of 53 students enrolled in the class took some portion of the test. Also, a group of 4 native
speakers (graduate assistants in the Spanish department) also took the test. Initially we had hoped to
run a control group with native speakers. However, since we were not able to obtain a group of native
speakers large enough to establish a comparison, we decided to exclude native speakers from the
present research. Thus, out of the 53 undergraduate students who took some part of the test, we
eliminated the results from the 6 undergraduate native speakers. We also eliminated the results from
the 8 students who missed the pre or the post-test.5 In short, the results that we are reporting in this
paper are the results obtained from the 39 nonnative speakers who took both the pre and the post-test.
3.2. Stimuli
The stimuli consisted of 34 words pronounced in one of three different ways. In one version, the
pronunciation was intended to be native-like. In a second version, consonant sounds were pronounced
in an English-like manner, and in the third version both consonants and vowels were English-like.6
Thus, there was a total of 102 items. Some of the targeted sounds are: fricative [β,ð,ɣ], flap [ɾ], trill [r],
un-aspirated [p, t, k], final [l], presence of [v] and [h], diphthongized vowels and [ə]. The words were
recorded by a non-native speaker of Spanish who is a linguistically trained speaker of English with the
ability to produce native-sounding Spanish words. All words were evaluated by native speakers: a
group of native speakers went over the test and we used those results to verify the accuracy of the
stimuli.
3.3. Method
The test was administered through the university course management system as part of the class
requirements. The students were asked to go to a specific computer lab on campus where they would
log on to the class web page.7 There they would find a link to the test. From the students’ perspective,
this test was just an activity that was graded for completion only. At the end of the semester, the
students were asked to give consent to use their responses in the present study.
The test was preceded by instructions regarding the test and a brief background questionnaire
whose main purpose was to identify native speakers of Spanish. The instructions that the students were
given are the following. (A Spanish version was also included.)
4
The instructor of this class is a native speaker from Spain.
Two of these excluded subjects actually took both the pre- and the post-test but had technical problems during
the post-test (they could not hear the stimuli) and answered randomly.
6
As pointed out by an anonymous reviewer, there is a logical fourth possibility: words pronounced with Englishlike vowels but Spanish-like consonants. We did try to produce this type of stimuli but we were not successful.
7
The access to the test was restricted to one particular computer lab on campus through an IP filter. If the subjects
tried to access it from anywhere else, access was denied.
5
238
(4) Experiment instructions
In the following, you will see a series of written Spanish words and you will hear a voice
that reads each word twice. After listening to the reading, you will have to evaluate the
pronunciation of each word on a scale from 5 (perfect, native like) to 1 (very bad). Before
that, you will have to answer a few questions about yourself.
5 Perfect / Native like -- Perfecto / Como un hablante nativo
4.
3.
2.
1. Very bad -- Muy mal
After reading the instructions and completing the background questionnaire, the actual test began.
A word appeared on the screen and a recording was played twice. The subjects had to rate the
recording and after doing that the new word / recording would appear. The order of presentation of the
102 items was randomized for each subject. The test was administered twice, once towards the
beginning (the pre-test) and once towards the end of the semester (the post-test).
4. Results and analysis
For illustration purposes, we will describe one of the words contained in the stimuli and a
selection of the subjects’ response for that word. One of the 34 words included in the stimuli was
“hubo”. This word was recorded in three different ways. Version A was native-like: approximant “b”
[β] and Spanish vowels. Version B had an incorrect occlusive “b” [b], but still with Spanish vowels.
Finally, version C not only had occlusive “b” [b] but a diphthongized final “o” [ou] also. The following
table includes the responses from the first ten subjects and the averages for all the 39 subjects.
(5) Sample answers and averages for “hubo”
Version A
[ú.βo]
Native-like consonants
and vowels
Version B
[ú.bo]
English-like, stop “b”
Version C
[ú.bou]
English like, stop “b” and
diphthongized final “-o”
Subject
pretest
posttest
Diff
pretest
posttest
Diff
pretest
posttest
Diff
1
5
5
0
3
2
1
1
1
0
2
4
5
-1
4
4
0
2
3
-1
3
5
5
0
2
4
-2
1
1
0
4
5
5
0
3
1
2
4
1
3
5
5
5
0
4
4
0
4
2
2
6
5
5
0
3
3
0
3
2
1
7
5
5
0
5
2
3
4
1
3
8
5
5
0
5
3
2
3
3
0
9
5
5
0
4
3
1
3
2
1
10
5
4
1
4
3
1
2
2
0
4.846
4.872
-0.026
3.692
2.769
0.923
2.410
1.718
0.692
⋮
Average
for all 39
Subjects
The data were averaged across the students’ responses on the pre-test and post-test for each of the
three versions (A, B, and C). Recall that group A contained both native-like consonants and native-like
vowels, B had generally native-like vowels but English-like consonants, and C was composed of
words with English-like consonants and vowels. The averages for the A, B, and C words in the preand post-test are shown below, along with the differences between the two tests and p-values for the
differences as determined by a paired t-test.
239
(6) Averages for all words by group type
Pre-test
Average
Post-test
Average
Difference
p-value
Group A
4.52
4.53
-0.01
0.811
Group B
2.58
2.12
0.46
0.000
Group C
1.67
1.39
0.22
0.000
The first item of interest is that the students appear to make a clear differentiation between groups
A, B, and C even in the pre-test, and this distinction also appears in the post-test. Words in group A
were assigned an average value of 4.52, very close to the ideal score of 5 for native-like productions.
Words in group B received a 2.58, indicating a middle value, while words in group C were given an
average value of 1.67, which is reasonably close to the value of 1 that words with a completely
English-like value should have received.
In terms of change in the students’ scores over the semester, it is clear that there is no
improvement in the native-like words from group A. However, the average started at 4.52, close to the
expected rating of 5 for native-like Spanish. Most of the A cases require no discussion since there is
little or no change; one exception in the A group will be discussed in detail later. There is a change in
the B and C words, which a paired t-test showed to be statistically significant. The results show that
students gave the words higher ratings at the beginning of the semester than at the end, indicating that
they realized at the end that the recordings were more English-like. Although we could use the
comparison between the groups B and C to try to figure out the impact that the English-like vowels
(diphthongized, long vowels, schwas) had in the ratings of each member of the group C, we will
exclude from the discussion the results of group C. We cannot directly infer anything from the
students’ reactions to the words in group C, since the vowels and consonants were both non-native-like,
and thus the students could have been reacting to either. Overall, the words in group B tended to show
the largest differences between the pre-test and post-test, showing a change in students’ reactions to
English-like consonants.
In what follows we describe the word types that we tested, giving examples for each type of
word.8
(7) Types of words that were tested
Word
These words
Some L2 learners…
Type
contain…
Z
the grapheme “z” pronounce the words
with a [z]
BDG
intervocalic /bdg/ fail to produce the
approximant allophone
found in Spanish
Final L word-final “l”
use the velarized “l” [ɫ]
8
H
the grapheme “h”
V
the grapheme “v”
PTK
/ptk/
R / RR
the flap [ɾ] or
alveolar trill [r]
in word-final position
pronounce the “h”,
forgetting that it is silent
in Spanish
pronounce the words
with a [v]
pronounce them
aspirated, as in English
pronounce them as an
alveolar approximant [ɹ],
like in English
Stimuli
“caza”and
“paz”
e.g. “hubo”,
“oda”,
“abogado”
e.g. “sal”,
“cual”
e.g. “mohoso”,
“bahía
e.g. “ave”,
“uva”
e.g. “taco”,
“pato”
e.g. “ara”,
“arra”
Stimulus
Versions
A: [ká.sa]
B: [ká.za]
A: [ú.βo]
B: [ú.bo]
A: [sal]
B: [saɫ]
A: [ba.í.a]
B: [ba.hí.a]
A: [ú.βa]
B: [ú.va]
A: [tá.ko]
B: [thá.ko]
A: [á.ɾa], [á.ra]
respectively
B: [á.ɹa] for both
The Z-, H- and V-words are different from the rest in that the subjects' knowledge of the grapheme-phoneme
relation is being tested. We thank an anonymous reviewer for bringing up this issue.
240







Z words: These are words that have the grapheme “z”. This type of word is problematic for
learners because they tend to pronounce them with a [z]. We tested two examples: “caza” y
“paz”. For “caza”, the A version was [ká.sa] and the B version was [ká.za].
BDG words: These are words that contain intervocalic /bdg/. These types of words are
problematic for the students because they normally fail to produce the approximant allophone
found in Spanish. Some of the examples that we tested were “hubo”, “oda” and “abogado”.
See above for the different versions of “hubo”.
Final L words: There are words with final “l”. Frequently students use the velarized “l” [ɫ] in
word-final position. Some of the words that we tested are “sal” and “cual”. The A version for
“sal” would be [sal] and the B version would be [saɫ].
H words: These are words that contain the grapheme “h”. Some students tend to pronounce it
forgetting that “h” in Spanish is silent. Some words that we tested are “mohoso” y “bahía”.
The A version for the latter would be: [ba.í.a] and the B version [ba.hí.a].
V words: These are words that contain the grapheme “v”. Some students pronounce these as a
[v]. Some words that we tested are “ave” y “uva”. The A version for the latter would be [ú.βa]
and the B version [ú.va].
PTK words: These are words that contain /ptk/. Some students pronounce them aspirated as in
English. Some of the words that we tested are “pato” y “taco”.
R / RR words: These are words that contain the flap [ɾ] or the alveolar trill [r]. Some students
pronounce them as an alveolar approximant [ɹ], like in English. Some of the words that we
tested are “ara” and “arra”. The A versions were [á.ɾa] and [á.ra] respectively, and the B
version was [á.ɹa] for both.
The following table summarizes the results for the word types that showed a statistically
significant difference between the pre- and the post-test.9 The results by word type were obtained in
the same manner as for the general analysis of groups A, B, and C.
(8)
Word types that showed a statistically significant difference (p < 0.05)
Z
BDG
-L
H
V
9
Group A
Group B
Group C
Group A
Group B
Group C
Group A
Group B
Group C
Group A
Group B
Group C
Pre-test
Average
4.14
2.24
1.37
4.69
3.16
2.20
4.45
2.00
1.53
4.45
2.91
1.83
Post-test
Average
4.44
1.55
1.18
4.81
2.43
1.67
4.41
1.54
1.28
4.56
2.37
1.36
Group A
Group B
Group C
3.78
3.04
1.97
3.39
1.58
1.30
Difference
p-value
-0.30
0.69
0.19
-0.12
0.73
0.53
0.04
0.46
0.24
-0.12
0.54
0.47
0.036
0.000
0.007
0.015
0.000
0.000
0.727 (ns)
0.000
0.006
N/A10
0.001
0.000
0.40
1.46
0.68
0.016
0.000
0.000
We have omitted from the discussion and will not report in this paper the data regarding words that could not fit
in exactly one of the types that we just described. Thus, we disregard the data for words like “acaba” which in the
B version has both an aspirated [kh] and an occlusive [b], or “cascabel” which in the B version has an occlusive
[b] and a final velarized [ɫ].
10
The paired t test could not be run for this particular set of data because of the distribution of the data. A
Wilcoxon Signed-Rank Test did not find the different statistically significant (p(2-tail)=0.3173).
241
Z WORDS: In Latin American Spanish, the grapheme “z” generally represents an unvoiced
alveolar fricative, and the voiced alveolar fricative occurs only as an allophone of the /s/ before voiced
consonants, so it is improper to pronounce words like “caza” as [ká.za]. This is unlike English, which
uses the “z” to represent the voiced alveolar fricative. In light of this difference, it is understandable
that students gave a rating of 4.14 to the productions of “caza” as [ká.sa]. This is a bit low when
compared to the overall average of 4.52 for the A group as a whole, and might indicate a lack of
awareness of the difference between the English Z and the Spanish Z. For group B, the students gave a
2.24 to productions of the Z with a voiced sibilant in the pre-test (for example, a word like “paz” was
pronounced as [paz] in group B, in contrast to [pas], which would be more typical in general Latin
American Spanish). In the post-test, this score fell by 0.69, to a 1.55, indicating a clear opinion that
[paz] is not a native-like production. From this, we infer that some students benefited from a
discussion of the differences between the Z in English and Spanish.
BDG WORDS: A similar pattern occurred for the voiced stops /b, d, g/. These phonemes differ
from the English in that they are only realized as stops after a pause or nasal, or in the case of /d/, after
an [l]. Intervocalically, they are realized as the approximants [β, ð, ɣ]. Thus, to maintain our pattern of
incorrectly produced consonants in group B, students heard words like “abogado” pronounced with
intervocalic stops, as in [a.bo.ɡá.do], rather than the appropriate intervocalic approximates in group A,
as in [a.βo.ɣá.ðo]. Students originally gave the group B production with stops an average rating of 3.16,
but lowered the score to a 2.43 in the posttest. This lends strong support to the idea that students
benefited from the discussion of the difference between /bdg/ in English and Spanish.
FINAL L WORDS: The data for the final L words show generally the same pattern as the data
overall: there was no sign of a significant change in the native-like words, while the non-native-like
words were initially rated somewhat high and received lower scores in the posttest. However, there
was a significant change in the B (and C group). Remember that there is an important difference
between Spanish and English for final “l.” While the word-medial and word initial “l” is the same in
both languages, English contains the word-final “dark l.” In English words like “wonderful,” the final
/l/ is velarized (and the tip of the tongue might not be touching the alveolar ridge), whereas Spanish
words like “cual” contain the same sound as the word-medial “l”. Final L words in the B group
(pronounced with a dark “l”: [saɫ], [kwaɫ]) were initially given a rating of 2.00 By the end of the
semester, this rating dropped significantly to 1.54.
H WORDS: The data for H words also show the same pattern as the data overall: there was no
sign of a significant change in the native-like words, while the non-native-like words were initially
rated somewhat high and received lower scores in the posttest. The Spanish “h” differs from English in
that in Spanish the grapheme “h” is not pronounced. The H words in the B group contained an
English-like [h]. Thus, for the word “bahía,” the stimuli included an [h], as in [ba.hí.a] rather than the
Spanish-like [ba.í.a]. Subjects gave these words with [h] an average score of 2.91 at the beginning of
the class, and a score of 2.37 at the end. Although a 2.37 is somewhat high considering that the
students were hearing a sound that should have been entirely silent, this is an improvement.
V WORDS: The V was an interesting case for two reasons, the first of which was the large
change in students’ judgments of it from the beginning and end of the semester. For group B, students
heard productions of words like “ave” as [á.ve], with the English voiced labiodental fricative, which is
not used in Spanish. At the beginning of the semester, students rated this at a 3.04, a relatively high
score, which plummeted to a 1.58 at the end of the semester. From this we infer that by the end of the
semester the students differentiated between the voiced bilabial and voiced labiodental fricatives,
recognizing that [á.ve] is not a native-like production.
The other notable feature of the data for the V was its remarkably low scores in group A on both
tests. Words such as “ave” and “uva” in group A obtained a score of 3.78, much lower than the rest of
the word types in group A, and the average fell to 3.39 in the post-test. This represents a significant
change, which puzzled us because we had expected the students to rate words from group A closer to a
five after the class due to their pronunciation with a bilabial approximant. It is important to note that
subjects did not show an unusual response to the same bilabial approximant when the words shown
contained the letter “b” rather than a “v.” One possible explanation for this unexpected result is that the
subjects saw a “v” and that made them “hear” a [v], which they know is not a good option in Spanish.
The interference of the letter “v” in the pronunciation of /b/ has already been noted by Zampini (1994).
242
So far we have looked at word types that showed a significant difference between the pre and
post-test. Now let’s turn our attention to those word types that didn’t.
(9) Word types that did not show a statistically significant difference
PTK
Group A
Group B
Group C
Pre-test
Average
4.59
2.39
1.48
Post-test
Average
4.46
2.33
1.41
RR
Group A
Group B
Group C
4.81
1.45
1.05
4.82
1.40
1.13
-0.01
0.05
-0.08
R
Group A
Group B
Group C
4.37
1.72
1.40
4.32
1.55
1.28
0.05
0.17
0.12
Difference
P-value
0.12
0.06
0.07
0.170 (ns/r=29)
0.342 (ns/r=30)
*0.039 (ns/r=26)
0.85
0.62
0.18
0.59
0.20
0.11
* = statistically significant
PTK WORDS: In the PTK words, there was no significant difference between the pre- and posttest in the A group (native-like) or the B group (aspirated consonants [ph, th, kh] but native-like vowels).
There was a significant difference in the C group, but remember that in this group the vowels are also
pronounced non-native-like.11 In other words, for a word like “taco,” there was no difference between
the pre and the post-test for the A version ([tá.ko]) nor in the B version ([thá.ko]). The only significant
difference was for the C version ([thá.kou]) with a final diphthongized vowel. Since the presence of the
aspirated “t” [th] did not make a difference in the B version, the difference in the C version is most
likely due to the diphthongized vowel, not the aspirated consonant.
R / RR WORDS: Finally, there was no significant difference between the pre and the post-test in
any of the versions of the R / RR words. Thus, using as an example the word “arra”, the A version
[á.ra] was already identified as the right pronunciation by most subjects in the pre-test (4.81, fairly
close to the highest possible value of 5), whereas the B version ([á.ɹa]) and the C version ([á.ɹə]) were
already given a fairly low score in the pre-test: 1.45 and 1.05, respectively.
To summarize our results, we saw significant progress toward the expected ratings between the
beginning and the end of the semester. Most of the word types showed an improvement, the only
exception being PTK and R/RR words.
5. Discussion
Since there is a significant difference between the pre- and post-test in most of the types of words,
formal phonetics instruction is supported. In most of the cases, students’ ratings increased for the A
type of words (those pronounced native-like) and decreased for B and C type of words (those
pronounced “with an accent”). Thus, this study corroborates the idea that teaching formal phonetics is
effective in improving the linguistic competence of the students.
11
For the PTK words, the data followed a non-normal distribution and the t-test was not applicable, so we
conducted a Wilcoxon signed-rank test on the data.
243
There were only two types of words that showed no significant improvement: PTK and R / RR
words. The reason for the lack of improvement in the R / RR words is most likely due to the fact that
the subjects’ ratings of those words were already very good in the pre-test so there was not much room
for improvement to begin with.
As for the lack of improvement in the PTK words, an explanation is less obvious. It is true that the
general rating for these words is better than the average, but many types of words with similar or better
ratings in the pre-test still showed improvement in the post-test. Several possibilities to explain the
lack of improvement on PTK words come to mind: maybe the instruction on this particular topic might
not have been good enough; alternatively, the difference between aspirated and unaspirated stops in
English and Spanish is less salient than the other differences studied in this experiment.
Another important contribution of this study is further support for an underused approach to
studying phonological proficiency. Compared to production based strategies, the Grammaticality
Judgment Task strategy that we have used to determine the competence of our subjects is remarkably
simpler, and still seems to be doing the work of tracking the students’ progress. We believe that this
opens the door for a reliable and simple way of testing phonological linguistic competence.
The approach that we have developed is obviously limited since it only measures the ability of the
subjects to distinguish good and bad pronunciations in the target language. But on the other hand, it
gives us the opportunity of obtaining quantitative data of one aspect of the linguistic competence of the
learner. It might not be the most important aspect of the linguistic knowledge of the learner, but it is
still part of it.12
We believe that the limitations of this technique are compensated by its ease of application. Once
the experiment is set up, any number of subjects can participate without any extra time or energy on
the part of the experimenter. The process of analyzing subject responses is normally very tedious, time
consuming and prone to error. By simplifying this aspect of phonological assessment, we have a
testing procedure that is simple enough to be applied to large subject populations.
Finally, we would like to comment on the overall performance of the subjects in the pre-test. In
many cases, there was not a big difference between the pre-test and the post-test with the group A
words because the initial values in the pre-test were already very high. Based on this, one could argue
that students at this level already have sophisticated phonetic knowledge that allows them to
differentiate between native and non-native pronunciation. On the other hand, there is impressionistic
data that students at this level are far from being able to produce the pronunciations that they judge as
native-like. This would be in agreement with Altenberg’s (2005) findings: She found a clear contrast
between the ability of her subjects to judge illicit consonant clusters and their ability to produce those
clusters correctly. If it is true that intermediate students have a sophisticated knowledge of Spanish
pronunciation that is much more refined than their performance would lead us to expect, there are two
additional conclusions that could be drawn.
One is related to the way we teach formal phonetics. Since intermediate students fail to produce
certain sounds in a native-like manner, it is assumed that they don’t know these sounds, but this may
turn out to be inaccurate. We should favor those approaches of phonetics that rely on the learners’
knowledge and use that knowledge as a teaching tool. A recent and interesting approach along these
lines is Lord (2008) who uses peer review of student created podcasts.
The other conclusion is related to the way we think about the SLA process. Frequently it is taken
for granted that there is a fundamental difference between the acquisition of syntactic knowledge and
phonological knowledge.13 However, this might not be accurate. A cursory look at the literature seems
to indicate that L2 learners achieve different degrees of success at syntactic and phonological levels.
But it could be the case that this is due to the ways in which we are measuring their knowledge.
Frequently L2 syntactic knowledge is assessed with judgment tests such as the grammaticality
12
Following Altenberg (2005), we assume that measuring this type of knowledge is valuable even if it turns out
that it is not directly correlated to other components of the phonological knowledge of the learner.
13
For instance, Moyer (1999) says: “Anecdotally, it is common to point to nonnative speakers (NNSs) whose
linguistic abilities are apparently native, with the exception of pronunciation. Seliger (1978) and Long (1990) have
suggested separate critical periods for different levels of language.”
244
judgment test or the truth value judgment test, which do not require the subject to produce any
utterances. On the other hand, phonological knowledge is more frequently tested using production tests,
and rarely tested using grammaticality judgment tasks as in the present study. The implicit
phonological knowledge that students seem to have could be used to support the idea that there is UG
access in phonology.14 After all, the students are showing a clear ability, for instance, to distinguish
between stop [b] and approximant [β], even in the pre-test.
As tempting as the conclusion from the previous paragraph may be, the evidence might not be
completely compelling. The results are telling us that when the subjects have to decide whether [í.ba]
or [í.βa] are native pronunciations of “iba”, they overwhelmingly identify [í.βa] as a native
pronunciation. This could be for two reasons. One is that they have access to UG, and even though
there is no [b] / [β] contrast in English, access to UG enables them to figure out the difference between
[b] and [β]. The other reason is that they simply assume that the most exotic variant, in this case [í.βa],
is the right one; after all, it is the one that sounds less English-like.
Although these conclusions are interesting, there is no doubt that future research is needed to
validate the findings in this study. For instance, to test the validity of the claims and arguments in the
previous two paragraphs, a similar test should be run with speakers that have little or no knowledge of
Spanish. There are also a few things in the test itself that could be changed. It would probably be a
better idea to use made-up words in the stimuli to avoid any type of familiarity effects. Each item in
the stimuli should exclusively target one particular sound. A control group of subjects not enrolled in a
pronunciation course should also be used. In the present study we saw a clear difference between the
pre-test and the post-test and we attributed this difference to the phonetics / pronunciation class they
were enrolled in. However, we have not ruled out the possibility that the improvement could have been
achieved by taking any other third year Spanish class. Having a control group would help us better
understand the impact of a phonetics class.
To conclude, we wanted to test whether a phonetics class at the intermediate level helps students
improve their knowledge of the pronunciation of Spanish. For that purpose, we designed an
experiment based on a grammaticality judgment task. Our results indicate that formal phonetics
instruction does have a positive effect. Also, our experiment reveals that our subjects have clear
intuitions regarding what a native pronunciation should be like, even though they are far from using it
in their production. If this is true, there could be an argument here for UG access. But before we
continue that line of reasoning, more research needs to be conducted to confirm these results. Finally,
we have developed a testing strategy that is remarkably simple, even if limited, which might open the
door to further study of L2 phonology.
References
Altenberg, E. P. 2005. The judgment, perception, and production of consonant clusters in a second language.
IRAL-International Review of Applied Linguistics in Language Teaching 43.1: 53-80.
Cenoz, J., and L. G. Lecumberri. 1999. The effect of training on the discrimination of English vowels. IRALInternational Review of Applied Linguistics in Language Teaching 37.4: 261-275.
Chaudron, C. 2003. Data collection in SLA research. In The Handbook of Second Language Acquisition, eds. C. J.
Doughty & M. H. Long, 762–828. Malden, MA: Blackwell.
Colantoni, L. and J. Steele. 2006. Native-like attainment in the L2 acquisition of Spanish stop-liquid clusters. In
Selected Proceedings of the 7th Conference on the Acquisition of Spanish and Portuguese as First and
Second Languages, eds. C. A. Klee and T. L. Face, 59-73. Somerville, MA: Cascadilla Proceedings Project.
Elliott, A. R. 1995. Foreign language phonology: Field independence, attitude, and the success of formal
instruction in Spanish pronunciation. The Modern Language Journal 79.4: 530-542.
Elliott, A. R. 2003. Phonology: Staking out the territory at the turn of the century: Integrating phonological theory,
research, and the effect of formal instruction on pronunciation in the acquisition of Spanish as a second
language. In Spanish Second Language Acquisition: State of the Science, eds. B. A. Lafford and M. R.
Salaberry, 19-46. Georgetown University Press.
Face, T. L. 2006. Intervocalic rhotic pronunciation by adult learners of Spanish as a second language. In Selected
Proceedings of the 7th Conference on the Acquisition of Spanish and Portuguese as First and Second
Languages, eds. C. A. Klee and T. L. Face, 47-58. Somerville, MA: Cascadilla Proceedings Project.
14
Grenon (2006) also supports the idea that there is UG access in L2 acquisition of phonology using an
experiment based on perception.
245
Flege, J. E. 2003. Assessing constraints on second language segmental production and perception. In Phonetics
and Phonology in Language Comprehension and Production, Differences and Similarities, eds. N. O.
Schiller and A. Meyer, 319-355. Berlin: Mouton de Gruyter.
Flege, J. E., and I. R. A. MacKay. 2004. Perceiving vowels in a second language. Studies in Second Language
Acquisition 26.1: 1-34.
Gaff, J. and S. Loewen. 2007. Teaching Spanish pronunciation: Does it work in the long run? Paper presented at
the Hispanic Linguistics Symposium at the University of Texas at San Antonio.
González-Bueno, M. 1997. The effects of formal instruction on the acquisition of Spanish stop consonants. In
Contemporary Perspectives on the Acquisition of Spanish, eds. W. R. Glass and A.T. Pérez-Leroux, Vol. 2, :
57–75. Somerville, Massachusetts, Cascadilla Press.
Grenon, I. 2006. Adults still have direct access to UG: Evidence from the perception of a non-native feature
contrast. In Proceedings of the 8th Generative Approaches to Second Language Acquisition Conference
(GASLA 2006), eds. M. G. O’Brien, C. Shea, and J. Archibald, 51-62. Somerville, MA: Cascadilla
Proceedings Project.
Guitart, J. M. 2004. Sonido y sentido: Teoría y práctica de la pronunciación del español. Georgetown Studies in
Spanish Linguistics. Washington, D.C.: Georgetown University Press.
Hammond, R. M. 2001. The Sounds of Spanish: Analysis and Application (with Special Reference to American
English). Somerville, MA: Cascadilla Press.
Long, M. H. 1990. Maturational constraints on language development. Studies in Second Language Acquisition
12.3: 251-85.
Lord, G. 2005. (How) Can we teach foreign language pronunciation? On the effects of a Spanish phonetics course.
Hispania 88.3: 557-67.
Lord, G. 2008. Podcasting communities and second language pronunciation. Foreign Language Annals 41.2: 364379.
Moyer, A. 1999. Ultimate attainment in L2 phonology. Studies in Second Language Acquisition 21.1: 81-108.
Nibert, Holly J. 2006. The acquisition of the phrase accent by beginning adult learners of Spanish as a second
language. In Selected Proceedings of the 2nd Conference on Laboratory Approaches to Spanish Phonetics
and Phonology, ed. M. Díaz-Campos, 131-148. Somerville, MA: Cascadilla Proceedings Project.
Pennington, M. C., and J. C. Richards. 1986. Pronunciation revisited. TESOL Quarterly 20.2: 207-225.
Pérez-Leroux, A. T., M. Majzlanova, and J. Sánchez-Naranjo. 2003. The light flashed until dawn: the L2
acquisition of the semantic mappings of Spanish tenses and problems of s-selection and compositionality. In
Proceedings of the 6th Generative Approaches to Second Language Acquisition Conference, eds. J. M.
Liceras, H. Zobl and H. Goodluck, 227-237. Somerville, MA: Cascadilla Proceedings Project.
Piske, T., I. R. A. MacKay, and J. E. Flege. 2001. Factors affecting degree of foreign accent in an L2: A review.
Journal of Phonetics 29.2: 191-215.
Quilis, A., and J. A. Fernández. 2003. Curso de Fonética y Fonología Españolas para Estudiantes
Angloamericanos. Madrid: Consejo Superior de Investigaciones Científicas, Instituto de Filología.
Sakow, M., and J.C. McNutt. 1993. Perception of /r/ by native speakers of Japanese and Korean: Internal and
external perception. IRAL-International Review of Applied Linguistics in Language Teaching 31.1: 46-53.
Schwegler, A., J. Kempff, and R. Barrutia. 2007. Fonética y Fonología Españolas. Hoboken, NJ: John Wiley &
Sons.
Seliger, H. W. 1978. Implications of a multiple critical period hypothesis for second language learning. Second
Language Acquisition Research: Issues and Implications: 11-19.
Sorace, A. 1996: The use of acceptability judgments in SLA research. In Handbook of Second Language
Acquisition, eds. W.C. Ritchie and T.K. Bhatia, 375-409. San Diego, CA: Academic Press.
Spicher, L. L., F. M. Sweeney, and R. Pelayo Coutiño. 2007. La Voz: Fonética y Fonología Españolas. Upper
Saddle River, N.J.: Pearson, Prentice Hall.
Stokes, J. D. 2005. ¡Qué Bien Suena!: Mastering Spanish Phonetics and Phonology. Boston, MA: Houghton
Mifflin.
Sutter, R.W. 1976. Predictors of pronunciation accuracy in second language learning. Language Learning 26.2:
233-253.
Terrell, T. 1989. Teaching Spanish pronunciation in a communicative approach. In American Spanish
Pronunciation: Theoretical and Applied Perspectives, eds. P. C. Bjarkman and R. M. Hammond, 196-214.
Washington DC: Georgetown University Press
Teschner, R. V. 2000. Camino Oral: Fonética, Fonología y Práctica de los Sonidos del Español. Boston, MA:
McGraw Hill.
Zampini, M. L. 1994. The role of native language transfer and task formality in the acquisition of Spanish
spirantization. Hispania 77.3: 470-81.
Zampini, M. L. 1998. The relationship between production and perception of L2 Spanish stops. Texas Papers in
Foreign Language Education 3.3: 85-100.
Zyzik, E. 2006. A look at missing objects in L2 Spanish. In Selected Proceedings of the 9th Hispanic Linguistics
Symposium, eds. N. Sagarra and A. J. Toribio, 192-202. Somerville, MA: Cascadilla Proceedings Project.
Selected Proceedings of the
12th Hispanic Linguistics Symposium
edited by Claudia Borgonovo,
Manuel Español-Echevarría,
and Philippe Prévost
Cascadilla Proceedings Project
Somerville, MA
2010
Copyright information
Selected Proceedings of the 12th Hispanic Linguistics Symposium
© 2010 Cascadilla Proceedings Project, Somerville, MA. All rights reserved
ISBN 978-1-57473-440-9 library binding
A copyright notice for each paper is located at the bottom of the first page of the paper.
Reprints for course packs can be authorized by Cascadilla Proceedings Project.
Ordering information
Orders for the library binding edition are handled by Cascadilla Press.
To place an order, go to www.lingref.com or contact:
Cascadilla Press, P.O. Box 440355, Somerville, MA 02144, USA
phone: 1-617-776-2370, fax: 1-617-776-2271, e-mail: [email protected]
Web access and citation information
This entire proceedings can also be viewed on the web at www.lingref.com. Each paper has a unique document #
which can be added to citations to facilitate access. The document # should not replace the full citation.
This paper can be cited as:
Ausín, Adolfo and Megan Sutton. 2010. An L2 Pronunciation Judgment Task. In Selected Proceedings of the
12th Hispanic Linguistics Symposium, ed. Claudia Borgonovo, Manuel Español-Echevarría, and Philippe Prévost,
234-245. Somerville, MA: Cascadilla Proceedings Project. www.lingref.com, document #2420.